Global satellite observations have become a crucial data source for environmental and climate research, services and applications. While important findings can be done by analysing single data sets or singe overpass separately, enormous possibilities for research breakthroughs are expected by combining various data sets together. The uncertainty quantification of data fusion needs to take into account several data sources with various underlying assumptions and typically different resolutions, scales and temporal coverage.
This mini-session focuses on recent achievements in the spatio-temporal data fusion methodologies and applications in satellite remote sensing.
10:30
Large scale data fusion system of remote sensing and in situ observations by dimension reduction
Marko Laine | Finnish Meteorological Institute | Finland
Show details
Author:
Marko Laine | Finnish Meteorological Institute | Finland
To be able to efficiently utilize the available environmental observations from various sources such as Earth observing satellites, in situ measuring instruments we need spatio temporal data fusion methods. The key elements of uncertainty quantification in a data fusion system include the representativeness and uncertainties of the measurements and models as well as the natural variability of the phenomena of interest in different spatial and temporal scales, which will serve as a background information needed to gap the missing data.
Proper UQ can be done with Bayesian statistics and by hierarchical state space description of the data, the processes, and the parameters defining the data fusion system. Kalman smoother techniques allow practical tools to do retrospective multi dimensional time series analysis by dynamical linear models. However, there are computational challenges when the dimension of the state space is high. This talk will demonstrate a data fusion system that utilises different dimension reduction techniques. The system is used to estimate Chlorophyll concentrations in the Baltic Sea using both satellite and in situ measurements.
11:00
Land surface albedo observations from a constellation of geostationary satellites
Jessica Matthews | North Carolina State University | United States
Show details
Author:
Jessica Matthews | North Carolina State University | United States
The WMO-led activity on Sustained and Coordinated Processing of Environmental Satellite Data for Climate Monitoring (SCOPE-CM) provides the infrastructure to ensure a continuous and sustained generation of climate data records (CDR) from satellite data. SCOPE-CM represents an international partnership between operational space agencies to coordinate the generation of CDRs. The SCOPE-CM effort to generate a unique land surface albedo CDR involves 5 different geostationary satellite positions and approximately 3 decades of satellite data. NOAA’s National Centers for Environmental Information is producing albedo products from both GOES-E (75°W) and GOES-W (135°W). These are being merged with like products from EUMETSAT based on METEOSAT (0° and 63°E) and from JMA based on the Geostationary Meteorological Satellite System (140°E). All agencies have implemented a common physics-based retrieval that includes estimations of error on retrieved parameters describing surface angular anisotropy determined through the inversion of a radiative transfer model using multiple geostationary images collected over a day under different illumination conditions. In 2020, NOAA’s Climate Data Record Program plans to provide the Level 2 surface albedo product over North and South America for 1995-present. Future plans of the SCOPE-CM collaborative include fusion of all Level 2 satellite products into a Level 3 near-global gridded land surface albedo CDR.
11:30
- CANCELED - Model based pixel-level image fusion for remote sensing
Zhengyuan Zhu | Iowa State University | United States
Show details
Author:
Zhengyuan Zhu | Iowa State University | United States
In the context of remote sensing, one typically has access to multiple images of the same location taken by different instruments with potentially different spatial resolution and at various time frequency. It is often useful to integrate such images into a composite image for further analysis, and quantify the uncertainty in the resulting composite image. In this paper we develop a general framework to address the image fusion problem using spatial-temporal varying coefficient model, and apply it to the problem of fusing LandSat and MODIS imagery to provide composite images with high spatial and temporal resolution, and consistent uncertainty quantification of the fused images.
12:00
Anomalies as a tool in detecting hidden features in satellite data sets
Johanna Tamminen | Finnish Meteorological Institute | Finland
Show details
Authors:
Johanna Tamminen | Finnish Meteorological Institute | Finland
Janne Hakkarainen | Finnish Meteorological Institute | Finland
Satellite data sets include typically various sources of uncertainties that need to be taken into account when drawing conclusions based on the data. While random noise cancels out nicely when data is averaged, many other types of uncertainties can have spatial or temporal correlations. Moreover, satellite data sampling is typically not evenly distributed, but includes patterns, e.g., due to drifting orbit, solar angles or cloudiness, creating problems, in particular, when data have also seasonal cycles or trends. Utilizing anomalies instead of direct observations can be beneficial technical approach for getting around these uncertainties when, e.g., merging temporally the data. We discuss the anomaly techniques in general and demonstrate the power of the anomaly technique introduced by Hakkarainen et al. (2016) in detecting carbon dioxide emission signatures by using several years of NASA's OCO-2 satellite data.
Reference: J. Hakkarainen, I. Ialongo, and J. Tamminen. Direct space-based observations of anthropogenic CO2 emission areas from OCO-2. Geophysical Research Letters, 43(21):11,400–11,406, 2016.